An AI agent just bought its own research data over Lightning, then turned it into this.
Left: the raw market data it paid 5 sats for. No API key, no account.
Right: the chart it built from that exact data.
Total cost for the whole paper: 20 sats (~$0.01). Here's how 🧵
1
3
227
Most "AI research" either hallucinates the numbers, or depends on humans wiring up API keys and billing for every source.
This agent did neither. It discovered the data it needed and paid for it, per call.
1
32
All it needed was the Lightning Enable MCP server and a small Lightning balance.
No FRED account. No keys to rotate. No rate-limit headers. Just:
request → 402 Payment Required → pay a few sats → get structured data back.
1
45
Prompt:
"Write a well-researched, cited post on the yen carry-trade unwind. Use up to 25 sats from my wallet + FRED data on agent-commerce.store (via discover_api). Include charts + citations. Then use the journalist agent to write it."
1
25
Under the hood, per series:
discover_api → finds FRED on agent-commerce.store
GET /series/NIKKEI225/data → 402
pay 5 sats → receive data + preimage
Then it repeated that for USD/JPY, the VIX, and the US 10-year.
1
30
And because it used real data, it caught something a vibes-take would miss:
the biggest Nikkei/VIX extremes in the window were not the Aug-2024 carry unwind, they were the April-2025 tariff shock.
Different event. The data drew the line.
May 26, 2026 · 4:44 PM UTC
1
41
Receipt:
• 4 data calls
• 20 sats (~$0.01)
• 0 API keys
• 0 accounts
• 0 human approvals
This is what agentic commerce looks like: an AI paying for exactly what it needs, per call.
Full paper, charts, data, and prompt:
github.com/refined-element/l…
1
72
For anyone who wants to read the paper :
github.com/refined-element/l…
27